Mercury Removal from Natural Gas with Sulfur-Impregnated Carbon

In the early 1970s, trace mercury accumulating in the cryogenic recovery section of the Skikda LNG plant in Algeria caused the catastrophic failure of a heat exchanger. The mechanism was mercury attacking aluminium alloy 6061 tubes. That single failure is why every modern LNG and NGL plant specifies mercury at the inlet to its cold box, and why the number they specify — below 0.01 µg Hg/Nm³ — sits roughly three and a half orders of magnitude below the 30–40 µg/Nm³ that has historically been a typical high, and five orders below the Pacific Rim levels above 1,000 µg/Nm³.

Note the date carefully. The published sources say “early 1970s.” They do not name a year, and neither will we.

The short version

• Aluminium is the vulnerable metal. Mercury causes liquid metal embrittlement in aluminium alloys, producing crack initiation and propagation, most readily near welds. UOP records heat exchanger failures, plant shutdowns and fires among the historical consequences.
• The specification is below 0.01 µg Hg/Nm³ — 10 ng/Nm³, roughly 1 pptv — at the cryogenic inlet. Four independent sources state the same figure.
• The chemistry is one-way. Hg⁰ + S → HgS, analogous to cinnabar, with a sublimation temperature around 584 °C. This is chemisorption. The bed is changed out, never regenerated.
• Capacity does not track sulfur loading. DOE-funded work found 36 wt% sulfur gave 594 µg Hg/g at 164.4 m²/g BET, while ~10 wt% sulfur on a higher-temperature impregnation gave 2,376 µg Hg/g at 823.7 m²/g.
• Disposal is regulated by mercury content. The TCLP limit D009 is 0.2 mg/L; EPA’s land disposal restrictions set 260 mg/kg total mercury as the threshold above which the waste must be retorted or roasted under RMERC.

Why does mercury destroy a cryogenic heat exchanger?

Because brazed aluminium heat exchangers are made of exactly the metal mercury attacks. Aluminium alloys suffer liquid metal embrittlement on contact with mercury — a corrosive attack that initiates and propagates cracks, particularly in the proximity of a weld. At Skikda the alloy involved was 6061, and the mechanism is reported as most aggressive in combination with water near 0 °C, which is precisely the condition inside a cold box.

Pall describes LME as a form of corrosion leading to crack initiation and propagation primarily in equipment constructed from aluminium. UOP’s account of the historical record lists heat exchanger failures, plant shutdowns and fires. The peer-reviewed treatment is Coade and Coldham’s 2006 paper in the International Journal of Pressure Vessels and Piping on the interaction of mercury and aluminium in heat exchangers in a natural gas plant.

The engineering consequence is asymmetric: the carbon bed is cheap and the cold box is not. A mercury guard bed is insurance against an asset that cannot be repaired in place.

How much mercury is in natural gas?

Enough to matter, and highly variable. Published figures span:

Those are different metrics from different authors and are not directly comparable. What they agree on is the shape of the problem: mercury is present nearly everywhere at some level, and the spread between a benign field and a difficult one is four to five orders of magnitude.

A clean region-by-region table could not be verified, and we will not print one. Regional generalisations about mercury in gas circulate widely and trace back to secondary summaries. The only defensible design input is a measurement on your own gas.

What is the specification target?

Below 0.01 µg Hg/Nm³ at the inlet to the cryogenic section. That is 10 ng/Nm³, roughly 1 pptv. Four independent published sources state the same number: two Digital Refining process articles, an academic review in IJEAIS, and Pall’s mercury removal literature, which gives it as typically <10 ng/Nm³ (approximately 1 pptv).

It is worth registering how demanding that is. If your inlet gas carries 1,000 µg/Nm³ — inside the published Pacific Rim range — the guard bed is being asked to remove five orders of magnitude of mercury, continuously, with a cracked brazed-aluminium exchanger as the downstream failure mode.

Where does the guard bed go?

UOP ranks upstream of the amine and dryer units as the preferred guard bed position for most plants. The target at the cryogenic inlet is below 0.01 µg Hg/Nm³.

UOP sets out four options and ranks them:

1. Upstream of the amine and dryer units — preferred for most plants.
2. Inside the dryers — applicable to silver molecular sieve only.
3. On the regeneration gas.
4. Downstream of the dryers — not preferred.

The rationale for going upstream is that it minimises the total mercury present before there is any opportunity for mercury to migrate to various locations in the plant. Mercury that gets past the guard bed does not simply pass through; it partitions into liquids, plates onto surfaces and reappears later, including in places where personnel exposure and decommissioning become issues.

There is a hardware corollary. Pall recommends a liquid/gas coalescer and a gas particle filter ahead of the adsorption beds, on the straightforward grounds that liquid water and solids degrade adsorbent capacity. A guard bed sized on a clean-gas isotherm and fed wet, dirty gas will not deliver its rated life.

The chemistry, and why it is one-way

Sulfur-impregnated carbon does not adsorb mercury. It reacts with it.

Hg⁰ + S → HgS. The product is mercuric sulfide, analogous to the mineral cinnabar, a stable, low-volatility solid with a sublimation temperature of about 584 °C. That stability is the whole point of the process: once mercury is converted, it stays converted at any temperature the gas plant will ever see.

It also settles the end-of-life question. This is chemisorption, not physisorption. There is no reversible bond to break with steam or hot nitrogen, and thermal reactivation is explicitly unsuitable for carbons loaded with heavy metals or inorganics — mercury and lead are named. The bed is changed out. Any supplier discussion of “regenerating” a mercury carbon should be interrogated: what is being regenerated, and to what?

A second consequence follows. Elemental sulfur sublimes, so a sulfur-impregnated bed is temperature-sensitive in service: running one hot risks losing impregnant rather than gaining kinetics.

Capacity is not about sulfur loading

This is the single most useful technical fact in mercury guard bed design, and it is counter-intuitive: a carbon with more sulfur on it can hold far less mercury.

The evidence is a DOE-funded study by Vidic at the University of Pittsburgh (DE-FG22-96PC96212, March 1997), which tested sulfur-impregnated carbons at 55 µg/m³ inlet mercury and 140 °C:

The carbon with by far the most sulfur has by far the least capacity. Impregnation temperature, not sulfur loading, decides where the sulfur sits.

Read the second and fifth rows together. BPL-S-250 carries 3.6 times the sulfur of BPL-S-600 and delivers a quarter of the mercury capacity. Its BET area has collapsed to 164.4 m²/g from the 1,040 m²/g of the virgin BPL.

The mechanism is the impregnation temperature. Low-temperature impregnation deposits ring-structured S₈, a bulky allotrope that plugs the pore network and buries most of the sulfur where mercury cannot reach it. High-temperature impregnation deposits shorter sulfur chains, which preserve surface area and leave the sulfur distributed across accessible internal surface. The result in this dataset is roughly 54 times the mercury capacity of the commercial reference carbon, at the same nominal sulfur loading.

The report also gives the theoretical stoichiometric capacity for complete HgS formation on HGR: 0.607 g Hg/g of carbon. Every real carbon in the study falls far below that. Complete conversion of the deposited sulfur is not what happens in a bed.

The practical instruction that follows is short. A sulfur content on a datasheet is not a capacity, and it is not even a good proxy for one. Ask where the sulfur sits, or better, ask for a capacity measured at stated conditions.

Datasheet capacity versus field capacity

Two published numbers bracket the gap. Calgon’s HGR-P product bulletin states that mercury capacity on that carbon “can be as high as 30 % by weight.” Idaho National Laboratory, measuring sulfur-impregnated carbon in off-gas service, reported that the highest observed mercury concentrations — in the first bed — averaged up to 19 wt% of the virgin carbon mass.

Those are not contradictory. One is a vendor’s stated maximum, the other a field measurement at one site under one set of conditions. But the distance between them is the distance between a purchase decision and a change-out schedule. A bed sized against 30 % and delivering 19 % runs out roughly a third sooner than planned.

The INL work gives the surrounding conditions that make its number meaningful: bed temperature 120–125 °C, superficial velocity below 0.5 m/s, and an estimated mass transfer zone depth of 15–30 cm corresponding to roughly 0.5–1 second residence time. That is what a usable capacity statement looks like — a number with the conditions attached.

How sulfur carbon compares with the alternatives

Three technologies compete for this duty, and the honest comparison is about constraints rather than a capacity league table.

The wet-gas limitation on carbon has a specific cause: capillary condensation of water and heavier hydrocarbons in the micropores at positions upstream of molecular sieve drying. That is also why the coalescer and particle filter ahead of the bed are not optional.

The market has moved. Digital Refining records that carbon-based options have been replaced in many facilities by base and noble metal-promoted alternatives, both non-regenerative and regenerative. That is a published trend, and it is fair to state it.

What we will not state is a head-to-head capacity number between sulfur carbon and metal sulfide. No such comparison could be verified. Anyone offering you one should be asked which two products, at which inlet concentration, temperature and humidity, to which endpoint.

Disposal, which most articles skip

Spent mercury carbon is a waste stream with its own regulatory structure, and it belongs in the total cost of ownership. Two facts set its shape for a mercury buyer: EPA’s evaluation of the RMERC treatment standard puts the dividing line at 260 mg/kg total mercury, at or above which the waste falls in the high mercury subcategory and must be treated by RMERC — retorting or roasting to volatilise the mercury and condense it for recovery — and EPA’s own document identifies spent activated carbon among the mercury-bearing wastes retorted at those facilities. The waste codes, the treatment standards and the change-out arithmetic are set out in regenerate or change out.

Note also that Suracsh does not operate reactivation services in the United States, and mercury-loaded carbon would not be a candidate for them if we did.

Suracsh’s mercury grades

Two grades are published, both coconut shell with a chemical impregnant that converts mercury to mercuric sulfide.

Two things to say plainly about that table, both of which our own product data already notes.

The mercury loading is specified on the granular grade and is not stated on the pelletized sheet. Ask for it when specifying. A pellet grade selected for pressure drop should not cost you the one performance number you are buying the product for.

Neither sheet states the test conditions behind the loading figure. By the argument of this entire article — that capacity depends on inlet concentration, temperature, humidity and endpoint, and that a 30 % vendor maximum and a 19 % field average can both be true — a bare “min 20 %” is not yet a design input. It is a starting point for a conversation about conditions.

If you are specifying a guard bed, the questions to put to any supplier are the same: inlet mercury concentration and its variability, gas temperature, water content and whether the bed sits upstream or downstream of drying, required outlet, acceptable pressure drop, and change-out access. A quotation issued without those is a price, not a design.

The full specification for both mercury grades is published in HTML at suracsh.us, and every datasheet downloads without a form or an NDA. The contaminant-to-grade mapping is at /selection-guide/.

Sources

• Digital Refining — Mercury removal processes (Skikda, concentrations, specification)
• Digital Refining — Handling mercury in gas processing plants
• Digital Refining — Mercury treatment options for natural gas plants
• UOP Honeywell — Mercury removal from natural gas production, brochure
• Pall — Mercury removal literature
• Coade & Coldham, The interaction of mercury and aluminium in heat exchangers in a natural gas plant, Int. J. Pressure Vessels and Piping, 2006
• Ohio Lumex — Variability of mercury in natural gas extraction, transportation and processing
• IJEAIS — Mercury removal review
• Idaho National Laboratory — Sulfur-impregnated carbon mercury adsorber performance
• Vidic, DOE/University of Pittsburgh, DE-FG22-96PC96212, March 1997
• Calgon Carbon — HGR-P product bulletin
• Min, Ahmad, Kim, Oh & Lee, Adsorption Science & Technology 33(3):251–262, 2015
• General Carbon — Reactivated carbon, limitations for heavy-metal-loaded carbon
• 40 CFR 261.24 — Toxicity characteristic (TCLP) regulatory levels
• US EPA — Waste specific evaluation of the RMERC treatment standard

Questions